altera_dspbook.pdf
《altera_dspbook.pdf》由会员分享,可在线阅读,更多相关《altera_dspbook.pdf(8页珍藏版)》请在冰点文库上搜索。
1DigitalSignalProcessing:
APracticalGuideByMichaelParker,AlteraCorporationIntroductionThisbookisintendedforthosewhoworkinorprovidecomponentsforindustriesthatusedigitalsignalprocessing(DSP).Thereisawidevarietyofindustriesthatutilizethistechnology.WhiletheengineerswhoimplementapplicationsusingDSPmustbeveryfamiliarwiththetechnology,therearemanyotherswhocanbenefitfromabasicknowledgeofitsfundamentalprincipals,whichisthegoalofthisbook-toprovideabasictutorialonDSP.Mathematicswillbeminimizedandintuitiveunderstandingmaximized.Thisbookwillde-mistifymanydifficultconceptslikesampling,aliasing,imaginarynumbers,frequencyresponse,etc.,usingeasytounderstandexamples,whileprovidinganoverviewofthefunctionsandimplementationusedinseveralDSPintensiveapplications.ThisisnotablackboardofferingofequationsasanexplanationonDSP.Thereaderneedonlybecomfortablewithhighschoollevelmathskills.Keyconceptswillbeemphasized.AfteracompletereadingyouwilllikelybeabletotalkintelligentlywithothersinvolvedinDSPcentricindustriesandunderstandmanyofitsfundamentalconcepts.OK,letsgetstartedWhatisDSP?
DSPisperformingoperationsonadigitalsignalusingadigitalsemiconductordevice.Mostcommonly,multipliersandaddersareused.Ifyoucanmultiplyandadd,youcanunderstandthetechnology.Digitalcircuitshavebecomeprogressivelycheaperandfaster,aswellasduetotheinherentadvantagesofrepeatability,toleranceandconsistencythatdigitalcircuitshavecomparedtoanalogcircuits.Ifthesignalisnotinadigitalform,thenitmustfirstbyconvertedordigitized.Adevicecalledananalogtodigitalconverter(ADC)isused.Iftheoutputsignalneedstobeanalog,thenisconvertedbackusingadigitaltoanalogconverter(DAC).Aspromised,mathematicswillbeminimized,butcannotbeeliminatedaltogether.Somebasictrigonometryanduseofcomplexnumbersisunavoidable(anearlychapterintroducestheseusingsimpleexamples).Thereisalsooneappendixsectionwhereverybasiccalculusisused,butthisisnotessentialtotheoverallunderstanding.Asthis7partseriesisonlyintendedasanoverviewofthecompletebookanditschapters,youwilllikelyhavetoactuallypurchasethisbook(goodformeandthepublisher)togainacompleteunderstanding.ItismygoalinthisseriesofarticlestogiveaclearandconcisesummaryofthebookfollowingitsTableofContents.ThisfirstpartwillcoverChapters12Chapter1:
NumericalRepresentationToprocessasignaldigitally,itmustberepresentedinadigitalformat.Thereareanumberofdifferentwaystorepresentnumbers,andthisgreatlyaffectsboththeresultandtheamountofcircuitresourcesrequiredtoperformagivenoperation.ThischapterfocusesonimplementingDSPandyoudonotneedtounderstandDSPfundamentalconcepts.Digitalelectronicsoperateonbitsthatformbinarywords.Thebitscanberepresentedasbinary,decimal,octalorhexadecimalorotherform.Thesebinarynumberscanbeusedtorepresent“real”numbers.TherearetwobasictypesofarithmeticusedinDSP,floatingpointorfixedpoint.Fixedpointnumbershaveafixeddecimalpointaspartofthenumber(i.e.1234,12.34or0.1234).Floatingpointisanumberwithanexponent(i.e.1,200,000wouldbeexpressedas1.2x106).MostofthediscussionfocusesonfixedpointnumbersthataremostcommonlyfoundinDSPapplicationsandashortdiscussiononfloatingpointnumbers.DSPusessignednumbersthatarebothpositiveandnegativenumbers,buthowarenegativenumbersrepresented?
Insignedfixedpointarithmetic,thebinarynumberrepresentationsincludeasign,aradixordecimalpoint,andthemagnitude.Thesignindicateswhetherthenumberispositiveornegative,andtheradix(alsocalleddecimal)pointseparatestheintegerandfractionalpartsofthenumber.Thesignisnormallydeterminedbytheleftmost,ormostsignificantbit(MSB).Theconventionisazerousedforpositive,andonefornegative.Thereareseveralformatstorepresentnegativenumbers,butthealmostuniversalmethodisknownas“2scomplement”.Thismethodisdiscussedindetailinthebook.Fixedpointnumbersareusuallyrepresentedaseitherintegerorfractional.Anintegerrepresentation,thedecimalpointistotherightoftheleastsignificantbit(LSB),andthereisnofractionalpartinthenumber.Foran8bitnumber,therangewhichcanberepresentedisfrom-128to+127withincrementsof1.Infractionalrepresentation,thedecimalpointisoftenjusttotherightoftheMSB,whichisalsothesignbit.Foran8bitnumber,therangewhichcanberepresentedisfrom-1to+127/128(almost+1)withincrementsof1/128.Inthebook,thischapterpresentsnumeroustables,witheachgivingequivalentbinaryandhexadecimalnumbers.ThebookwillcoverIntegerFixedPointRepresentation,FractionalFixedPointRepresentation,andFloatingPointRepresentationindepth.Chapter2:
ComplexNumbersandExponentialsComplexnumbershaveprobablybeenforgottenbymostofus.TheyareimportantindigitalcommunicationsandDSP,soyoullneedtoresurrect3Acomplexnumberhasa“real”and“imaginary”part,andtheimaginarypartisthesquarerootofanegativenumber,whichisreallyanon-existentnumber.Soundweird,wellitstechnicallytrue,soitwillbeexaminedinamuchmoreintuitivewayinthebook.ComplexnumbersareneededtodefineatwodimensionalnumberplanetounderstandDSP.Thetraditionalnumberlineextendsfromplusinfinitytominusinfinity,alongasingleline.TorepresentmanyoftheconceptsinDSP,twodimensionsareneeded.Thisrequirestwoorthogonalaxes(i.e.thehorizontallineistherealnumberline,theverticallineistheimaginaryline).Allimaginarynumbersareprefacedby“j”,whicharedefinedasthe-1.ThisistheessenceofthiswholechapterasshowninFigure1.Figure1.AComplexNumberPlaneAnycomplexnumberZhasarealandimaginarypart,andisexpressedasX+jY,orjustX+jY.ThevalueofXandYforanypointisdeterminedbythedistanceonemusttravelinthedirectionofeachaxistoarriveatthepoint.Itcanalsobevisualizedasapointonthecomplexnumberplane,orasavectororiginatingattheoriginandterminatingatthepoint.Therehastobeawaytokeeptracktheverticalandhorizontalcomponents.Thatswherethe“j”comesin.ComplexAddition,SubtractionandMultiplicationexplanationsandexampleswillbegiveninthebook,aswellasin-depthoverviewsofPolarRepresentationandComplexMultiplicationusingPolarRepresentationexamples.Twopointswillbedefined,Z1andZ2.Z1=R1angle
(1)Z2=R2angle
(2)Z1Z2=(R1R2)angle(1+2)4Whatthismeansisthatwithanytwocomplexnumbers,themagnitude,ordistancefromtheorigintotheradius,getsmultipliedtogethertoformthenewmagnitude.Exampleswillbegiveninthebook.ComplexConjugateThisisthelastexamplethatwillbecoveredinthebook.Thisisaspecialcaseandwillbeexplainedwhy.Imaginarynumbersareusedtoformcomplexnumbers.Theyarereallynotsocomplex,andimaginaryisreallyaverymisleadingdescription.Whatwillbeexplainedistohowtocreateatwodimensionalnumberplaneanddefineasetofexpandedarithmeticrulestomanipulatethenumbersinit.Nextthecomplexexponentialwhichissimplytheunitcircle(radius=1)onthecomplexnumberplanewillbeexplained.Thelastpartofthischapterinvolvesmeasuringanglesinradians,whichisseeneverywhereinDSP.Theanglemeasurementinradiansisbasedupon,whichisanumberdefinedtohaveavalueofabout3.141592(itactuallyisanirrationalnumber,withinfinitenumberofdigits,like1/3=0.3333.).Ittakesexactly2radianstodescribeafullcircle.Thissameconceptwillbeexaminedlaterinsamplingtheory,whereeverythingtendstowraparoundorbehaveperiodically.Wecanvisualizethisastravelingeitherclockwise(negativerotation)orcounterclockwise(positiverotation)aroundthecircle.ThereisonemoreDSPconventiontobeawareof.Therealcomponent(Xwasusedearlier)itisusuallycalledthe“I”or“in-phase”component,andtheimaginarycomponent(usedY)isusuallyreferredtoasthe“Q”or“quadrature”component.InmanyDSPalgorithms,thedigitalsignalprocessingmustbeperformedsimultaneouslyonbothIandQdatastreams,whichsimplyrepresentsthesignalsmovement,overtime,withinthetwodimensionsofthecomplexnumberplane.Chapter3:
Sampling,AliasingandQuantizationThestartingpointtounderstandDSPissampling,anditsaffectonthesignalofinterest.TheADCmeasuresthesignalatrapidintervals,calledsamples.Adigitalsignalproportionaltotheamplitudeoftheanalogsignalatthatinstantisoutput.Ifasignalissampledveryfastcomparedtohowrapidlythesignalischanging,afairlyaccuratesamplerepresentationofthesignalisobtained,butifthesampleistooslow,adistortedversionofthesignalisseen.Figures2and3aresamplingsoftwodifferentsinusoidalsignals.Theslowermovingsignal(lowerfrequency)canberepresentedaccuratelywiththeindicatedsamplerate,butthefastermovingsignal(higherfrequency)isnotaccuratelyrepresentedbythesamplerate.Infact,itactuallyappearstobeaslowmoving(lowfrequency)signal,asindicatedbythedashedline.Thisshowstheimportanceofsampling“fast”enoughforagivenfrequency5Figure2.Samplingalowfrequencysignal(arrowsindicatesampleinstants)Figure3.Samplingahighfrequencysignal(samesampleinstants)Thedashedbluelineshowshowthesampledsignalwillappearifthesampledotsandsmoothoutthesignalareconnected.Noticethatsincetheactual(redsolidline)signalischangingsorapidlybetweensamplinginstants,thismovementisnotapparentinthesampledversionofthesignal.Thesampledversionactuallyappearstobealowerfrequencysignalthantheactual